Survey researchers have asserted several reasons why responses in cell phone interviews may be less accurate than responses in landline interviews. Somewhat surprisingly, studies comparing response quality from landline and cell phone interviews have found minimal differences. These studies were not entirely conclusive, however, because landline and cell RDD respondent samples are known to differ systematically with respect to more than just the device. It is possible that differences in sample composition confounded prior device comparisons. This study reduces that confound substantially by randomly assigning people to be interviewed on either their landline or their cell phone in a two-wave experimental design. The findings are generally consistent with the literature, showing no evidence of a device effect on most metrics. These results are encouraging for researchers designing surveys that feature cell phone data collection.